Fig. 1.

Outlines of the experimental set-ups. a Channel experiments. Nest and feeder were connected by a training channel; a test channel was arranged upwind in parallel and close to the training channel. Experimental animals were guided into either test or training channel by a switch door. Above the test channel, desert ants search behaviour around the assumed feeder position is illustrated schematically. The initial three turning points are indicated; the initial six turning points were evaluated, or just the first turning point. b Field experiments. The alley of black cylinder landmarks is shown as it extends from the nest entrance to the feeding site and 17 m further. Please note that the diagrams in a and b are not to scale, e.g. the nest-feeder axis is strongly compressed in b to accommodate the landmark alley of 27 m length, with only 0.5 m width; landmark alley is drawn to scale in Fig. 5a. c Icons used in the following figures to indicate experimental groups that were not manipulated and had normal eyes (left) and ants that had their ventral and lateral eye parts covered with opaque paint (right). The background picture shows a nest entrance, with a white 2 × 2 m recording grid painted on the desert floor